Argument Blogging

Simon Wells, Colin Gourlay, and Chris ReedSchool of ComputingUniversity of Dundee

{swells,cgourlay,chris}@computing.dundee.ac.uk

AbstractArgument Blogging is the process of harvestingtextual resources from the web and structuringthem into distributed argumentative dialogues. Thispaper introduces a prototype software system forperforming argument blogging and storing the re-sultant dialogues so that they can be analysed andreused.

1 IntroductionOnline argumentation is the process of engaging in and mak-ing use of arguments on the web and the online communityhas spontaneously demonstrated great interest in argument.This may have been spurred in part by the highly visible argu-ments with strong and explicit argumentative structure suchas those found in the Iraq Study Group Report 1. This in-terest has occured in a range of communities, from the non-expert communities such as users who wish to engage witha domain and find that arguing is a natural and intuitive in-teraction mechanism, and the developers who are building ar-gumentation tools because they recognise that argumentationcan be an engaging interaction mechanism, to the expert com-munities, mainly working in academia, who develop tools tosupport the capture, analysis, and sharing of argumentativedata.

Various online systems have been deployed to exploitthis interest such as convinceme.net2 and debatepedia.com3.These systems have enabled members of the public to en-gage with each other and to express their opinions using web-based interaction mechanisms. In convinceme.net the aim iscompetitive. The users attempt to accumulate votes, whichequate to points, with the aim of accumulating the most pointsand thereby becoming King of the Hill within a single de-bate. However, Debatepedia uses a Wikipedia style interfaceto collate large evidence sets to support a user in exploringand understanding a complex debate topic. What both ofthese systems have in common is that they both provide highquality Web 2.0 based interfaces and environments to sup-port user interaction. The interfaces arguably contribute to

1http://www.usip.org/isg/iraq study group report/report/1206/2http://www.convinceme.net3http://debatepedia.com

the construction of the broad userbases which underpin thesesystems but the underlying argumentation theory is often im-poverished, having small sets of moves which the users canmake and limited tools for interacting with the argumenta-tive content. Even where systems have not been designed toexplicitly support the dialectical nature of argumentative dia-logue, users will try to fit in a rudimentary yet intuitive argu-mentative structure. This occurs often in blog comments andin the BBC’s Have Your Say webpage4 in which users willoften manually copy and paste earlier posts into their ownresponse to specify the exact point to which they are respond-ing. Users obviously want a way to structure their interactionsand responses so that they can explore the arguments of otherswhilst making their own arguments explicit in relation. Suchcapabilities are not however explicitly supported by the cur-rent crop of weblog commenting or fora software. There hasalso been broad interest in the underlying formal representa-tions that support the widespread sharing and interchange ofargumentative resources in online systems. This is useful, notonly to support the development and deployment of persua-sive argumentation-based interfaces, but also to support moreadvanced online argumentation processing. Work towardsthis end has culminated in the Argument Interchange Format(AIF) [Chesnevar et al., 2006] used to record and share ar-gument resources and a foundational element of the proposedWorld Wide Argument Web (WWAW) [Rahwan et al., 2007].The juxtaposition of these two situations, end-users desiringmore structured interaction mechanisms with explicit supportfor argumentation, and the academic move towards an on-line network for sharing argumentation resources, suggestsan opportunity for new software tools that can support bothapproaches. It is this opportunity that is addressed in thispaper, the provision of an engaging new end-user interfacethat enables existing text-based online resources to be reused,structured and analysed in terms of their argumentative struc-ture.

2 Argument Blogging & the WWAWMuch as the world wide web (WWW) has provided a mech-anism for sharing information online between humans, andthe semantic web seeks to enable machines to understandand share information, the WWAW [Rahwan et al., 2007]

4http://news.bbc.co.uk/1/hi/talking point/default.stm

seeks to provide a similar mechanism for argumentation en-abling information to be stored on the Internet accordingto its underlying argumentative structure, and subsequentlyreused within a plethora of software applications. Oneway of interpreting the WWAW is therefore as a networkof argumentation-centered software tools that share a com-mon format for exchange of information between them. Ar-gument Blogging is so called because it takes the idea ofweb-blogging, posting insightful comments online in a user-driven, participatory fashion, and overlays a dialogue gameto guide user interaction and make explicit the argumentativerelationships between posts, comments, web sites and otheronline textual data. The aim is to produce a system that sup-ports the harvesting of existing text posted on the web andthe restructuring of those texts into argumentative dialoguestructures. The rationale for this pursuit is that there is a lotof argumentative interaction occuring on the web but thereare very few tools to enable these arguments to be capturedand resused.

3 Underlying TechnologiesArgument blogging is built upon a number of existing un-derlying technologies. These include standard web technolo-gies such as Apache HTTP web server5, Javascript6, JQuery7,Python8 and Django9, as well as argumentation-specific tech-nologies like the Argument Interchange Format (AIF), theAIF Database (AIFDB) and the Dialogue Game DescriptionLanguage (DGDL)10.

Argument Interchange Format (AIF) The Argument In-terchange Format (AIF) [Chesnevar et al., 2006] is a coreontology of argument-related concepts which has been for-mulated to provide flexibility in capturing a variety of ar-gumentation formalisms and schemes. The core ontologyis a directed graph containing two types of nodes, Infor-mation nodes (I-nodes) that represent passive informationand scheme nodes (S-Nodes) that capture the application ofschemes representing patterns of reasoning and inferential re-lationships between sets of I-Nodes. By providing a flexiblelanguage which supports a wide range of formalisms and canbe used to share structured analyses of arguments between awide range of end-user and back-end software tools, the AIFis thus a core tool that underpins the emerging WWAW.

AIF Database (AIFDB) The AIF Database (AIFDB) isa MySQL database that supports the storage of and subse-quent access to AIF resources. The AIFDB has been de-veloped to enable AIF documents, serialised in RDF, to beparsed, split into their constituent elements, stored, and sub-sequently reused by argumentation software. The interface

5http://httpd.apache.org/6http://www.ecma-international.org/publications/standards/Ecma-

262.htm7http://jquery.com/8http://www.python.org/9http://www.djangoproject.com/

10http://www.arg.computing.dundee.ac.uk/?page id=147

to the AIFDB is either through a regular bridge from a pro-gramming lanuage, such as Java, to the underlying MySQLdatabase, or via a RESTful web services interface that sup-ports the import and export of AIF document resources. TheAIFDB was developed separately to the argument bloggingsystem with the aim of creating a software tool that can beused to implement the WWAW, by supporting storage andreuse of AIF resources regardless of end-user and developerrequirements.

Dialogue Game Description Language (DGDL) Dialec-tical games [Hamblin, 1970] are a type of multi-player argu-mentative dialogue game that provide a mechanism for struc-tured communicative interaction whilst incorporating argu-mentative behaviours. However there have been very fewtools for working with these games and little agreement overhow they should best be described, shared, and reused. TheDialogue Game Description Language (DGDL) [Wells et al.,2009] is a domain specific language, in the sense of SQL fordatabase queries or LATEXfor typesetting, but in this case fordescribing the rules of dialectical and other dialogue gamesand provides a grammar for determining whether a game de-scription is syntactically correct and thus provides a foun-dation for new tools to support the future development andwider exploitation of dialogue games. The argument blog-ging system adopts the DGDL as an underlying technologythat supports the sharing of verifiable, syntactically correctinteraction protocols between online software tools.

4 Argument Blogging: ArchitectureThe argument blogging system uses a distributed client-serverarchitecture in which users can arbitrarily select sections ofargumentative text, published on the web, and link that intoa distributed dialogue stored on the WWAW. The overall ar-chitecture of the argument blogging system is illustrated inFigure 1 which shows two modes of interaction with the sys-tem, one as a resource gatherer, linking new argument re-sources into the WWAW via a JavaScript bookmarklet, andthe other as a transcript explorer, interacting with existingdialogues formed from previously gathered resources. Bothmodes of interaction are mediated by the argument bloggingserver which collates the argument resources and assemblesthem into cohesive dialogues which are then stored, as AIF,within the AIFDB, or visualised using the AIF argument vi-sualiser.

A disadvantage of this approach is that the server providesa point of failure for the sytem. However, whilst much ofthis functionality could be implemented using a combinationof the end user’s browser, and the argument resources in theweb, and the argument blogging system distributed across allof its users, the adoption of a centralised server has a key ad-vantage. The web is inherently dynamic and it is likely thatmany of the argument resources used to create distributed di-alogues using the argument blogging system will disappearover time which would lead to breaks in the chain of argu-ment if the user were to follow the dialogue from site to site.The alternative, using a central server, means that once re-sources are linked into a dialogue, the selected resource is

User

HTTP

Argument Blogging System

Python Scripts + Django Libraries

AIF DB

Resource Gathering

Transcript ExplorationHTTP

Flash/ActionscriptGraph Renderer

WWW

Argument Visualisation Tool

Argument Blogging Server

Figure 1: The Argument Blogging System Architecture

stored within the AIFDB and is thus available so long as theAIFDB is available. This avoids the problems that could ariseif the integrity of a chain of arguments distributed across theweb had to be maintained for the dialogue to be followed. Italso means that there is a repository from which argumen-tation resources, consisting of distributed online arguments,can be retrieved and analysed.

5 Argument Blogging: UsageThe interface to the argument blogging system has been madesimple and straightforward, to reduce possible barriers to par-ticipation, without compromising functionality. Currentlythere are two end-user modes of interaction; resource gath-ering and transcript exploration. Resource gathering is per-formed on text contained within any web page on the web.The process is simple, resource gatherers select a portion oftext within their web browser, then click upon a JavascriptBookmarklet, either stored in their toolbar or bookmarksmenu, which causes a dialogue box to be displayed as shownin Figure 2. The dialogue box offers the user a number ofchoices as to how they wish to respond with respect to theirselection. These choices constitute moves in a simple dia-logue game, specified using the DGDL, enabling the user toeither support, or refute a statement, or to attack an inferencebetween statements. The use of DGDL resources means thatthe interaction protocol describing the structure of the dia-logue between blogs can easily be altered to support differentargumentative behaviours.

Once a user has selected their response a dialogue is dis-played that contains a fragment of embeddable javascript. Forexample:<script type="text/javascript"src="http://134.36.36.189/core/quotes/55/badge.js"></script>

Users can select and copy this text and can paste it towherever their response will be located using an interactionmechanism similar to that used by mainstream websites likeYouTube11. Often the destination for this code is to either

11http://www.youtube.com

Figure 2: The Argument Blogging Bookmarklet.

a post or comment box within existing blog software, but itcan be pasted directly into the HTML of a users website orany other publically accessible site that allows and supportsoutbound links.

The root of a dialogue, the initial text that is selected andlinked into a dialogue via the bookmarklet, thus spurring adiscussion, cannot be altered directly because the author hasnot necessarily agreed to be part of the WWAW so is onlyconnected to the dialogue via a standard web link, however allsubsequent responses are as a result of users explicitly optinginto the system. Where users have opted into the argumentblogging system by publishing a script, a badge is displayedon the upper left corner of the page, shown in Figure 3, thatindicates that the page is a part of the WWAW. The WWAWbadge links back to the specific dialogue that the current pageis a part of hosted on the argument blogging server. Thebadge is, therefore, the first explicit, navigable link into theWWAW from an arbitrary web page hosting source content.

In addition to the resource gathering interaction there isalso the transcript exploration interaction. Following thebadge link leads the user to the transcript of the extended,distributed dialogue that the initial page is a part of. An ex-ample transcript is illustrated in Figure 5 which shows threeweb resources. The first resource, at the top of the page, is theroot of the dialogue and contains a link back to the originalpost where more context can be gained. For each responseto the initial resource, there is a link back to the previous re-source and a link to the original context of the current re-source. Each resource also indicates how it is related to theearlier resources in the dialogue, using the default game thisis restricted to supporting, refuting, or attacking the inferen-tial relationship between resources.

At the top of each transcript page is a link to enable a dia-logue to be exported in plain text format. This means that di-alogues can be reused in different contexts, for example, thefollowing is a short dialogue transcript containing the threeresources from Figure 5 that have been extracted from the

Figure 3: WWAW system “opt-in” via website badge. This isdisplayed automatically once the javascript code is embeddedinto the webpage.

Figure 4: The transcript of a single distributed dialogue usingan “instant message” style visualisation of the sequence ofutterances in the dialogue.

system using the plaintext export mechanism:In addition to plain text export, any given dialogue tran-

script can also be exported as AIF through a web link fromthe top of the dialogue page, a process that is illustrated inFigure 5. This means that argument resources gathered us-ing the argument blogging interface can subsequently be con-sumed by other argumentation tools that understand AIF suchas Araucaria [Reed and Rowe, 2001], ArgDF [Rahwan et al.,

Figure 5: Text Export of a single dialogue.

2007], and Cohere [Buckingham Shum, 2008]. This meansthat the argument blogging software works as a new interfaceto the WWAW, enabling arbitrary text based resources to begathered and structured in terms of their argumentative rela-tions, and providing resources using a simple and naturalisticinterface that can subsequently be analysed and reused usingmore appropriate task-specific tools. Also at the top of eachtranscript page is a link to submit the argument to an onlineargument visualisation tool so that a diagrammatic represen-tation of the dialogue based upon the AIF file can be viewed.Whilst there are a number of tools that understand and useAIF, none have yet been developed to produce native visual-isations of AIF resources, they have all, thus far, reused theirown existing argument and dialogue diagramming techniquesusing the AIF merely as a conduit for sharing arguments. Theargument blogging system utilises a new online argument vi-sualisation tool12, illustrated in Figure 6, that renders AIFresources into standard AIF diagrams. In these diagrams I-Nodes are represented by boxes and S-Nodes by diamonds.Directed edges between these nodes are used to indicate rela-tionships between the nodes.

In addition to exploring an individual argument, users canfollow the view all dialogues link from the top of the dialoguetranscript. This leads to the full list of dialogues held in theargument blogging system as illustrated in Figure 7.

6 Conclusions & Further WorkThe work discussed here is only a prototype for an argumentblogging system that supports that harvesting of existing ar-

12http://www.arg.computing.dundee.ac.uk/?page id=143

Figure 6: Visualisation of the AIF record for a distributed on-line dialogue captured using the argument blogging system.

Figure 7: A list of dialogues currently indexed by the argu-ment blogging server.

gumentative resources from the web and their import into theWWAW. As a result there are a number of avenues for func-tional enhancement and further research. For example, theinterface to the Dialogue List would benefit from additionalfunctionality. Currently the dialogue list interface merelyadds each new dialogue to the end of a list. This needs tobe enhanced so that as well as getting the list of the latest di-alogues, users can retrieve dialogues via a search interface.As the volume of users increases and the number of dia-logue resources increases, it becomes imperative to provideflexible and efficient methods for accessing those resources.One design aim was to enable the lowest cost of entry to theWWAW without compromising functionailty and usability.This is why the initial prototype system uses a bookmarkletto gather resources, but relies heavily upon a server to collatethe resources. Reliance on the server could constitute both

a bottleneck in the system during heavy load and a signifi-cant point of failure for the entire system. For this reason twostrands of functional enhancements are planned. The first isprimarily an architectural enhancement, and the second is afunctional enhancement. Functionality can be increased byusing plugins for popular blogging software such as word-press13, or browsers such as Firefox14, and javascript applica-tions to provide alternative methods for users to interact withthe system at various levels of involvement. The wordpressplugin and javascript are designed to enable website ownersto opt-into the WWAW system by including argument blog-ging functionality directly within their site. Increased abilityto interact with and manage arguments can be introduced forend users through the adoption of a Firefox plugin. Furtherresearch is also required into the dialogue games used by thesystem. Whilst the adoption of the DGDL simplifies the jobof switching to alternative interaction protocols, it is not clearwhich protocols would be most appropriate for this kind ofinteraction, for example more comprehensive protocols mightlose some of the dialectical structure from the dialogue at theexpense of usability and immediacy. Ultimately, the systemdescribed herein is but a first step on the road towards a webof argument that enhances the functionality of the web, whilstproviding a valuable resource to argumentation researchers.

AcknowledgmentsDevelopment of the AIFDB and Argumentation visu-aliser used in this work was performed by Mark Snaith(m.snaith@dundee.ac.uk) as part of an undergraduate degreeproject.

References[Buckingham Shum, 2008] S Buckingham Shum. Cohere:

Towards web 2.0 argumentation. In 2nd Interna-tional Conference on Computational Models of Argument(COMMA 2008), 2008.

[Chesnevar et al., 2006] C. Chesnevar, J. McGinnis, S. Mod-gil, I. Rahwan, C. Reed, G. Simari, M. South,G. Vreeswijk, and S. Willmott. Towards an argumentinterchange format. Knowledge Engineering Review,21(4):293–316, 2006.

[Hamblin, 1970] C. L. Hamblin. Fallacies. Methuen and Co.Ltd, 1970.

[Rahwan et al., 2007] I. Rahwan, F. Zablith, and C. Reed.Laying the foundations for a world wide argument web.Artificial Intelligence, 171:897–921, 2007.

[Reed and Rowe, 2001] C. Reed and G. Rowe. Araucaria:Software for puzzles in argument diagramming and xml.Technical report, University Of Dundee, 2001.

[Wells et al., 2009] S. Wells, A. Ravenscroft, M. Singh, andC. Reed. Mapping persuasive dialogue games onto argu-mentation structures. In Symposium on Persuasive Tech-nology held at the AISB Convention in Edinburgh, 2009.

13http://wordpress.org/14http://www.mozilla.com/firefox/